[KJ] GCC4 cleanup question

[KJ] GCC4 cleanup question

[Chuck Short]

Hi,

I was wondering how would you fix warnings like the following:

net/ipv4/route.c: In function 'ip_route_input_slow':
net/ipv4/route.c:1770: warning: 'rth' may be used uninitialized in this 
functionnet/ipv4/route.c:1797: warning: 'rth' may be used uninitialized 
in this functionnet/ipv4/route.c: In function 'ip_route_output_slow':
net/ipv4/route.c:2270: warning: 'rth' may be used uninitialized in this 
functionnet/ipv4/route.c:2242: warning: 'rth' may be used uninitialized 
in this function

Thanks
chuck
_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
http://lists.osdl.org/mailman/listinfo/kernel-janitors

Re: [KJ] GCC4 cleanup question

[Bongani Hlope]

On Saturday 28 May 2005 06:00 am, Chuck Short wrote:
> Hi,
>
> I was wondering how would you fix warnings like the following:
>
> net/ipv4/route.c: In function 'ip_route_input_slow':
> net/ipv4/route.c:1770: warning: 'rth' may be used uninitialized in this
> functionnet/ipv4/route.c:1797: warning: 'rth' may be used uninitialized
> in this functionnet/ipv4/route.c: In function 'ip_route_output_slow':
> net/ipv4/route.c:2270: warning: 'rth' may be used uninitialized in this
> functionnet/ipv4/route.c:2242: warning: 'rth' may be used uninitialized
> in this function
>
> Thanks
> chuck

I don't have GCC4, the first rth is initialised inside the __mkroute_input 
function and the second one is initialised in the __mkroute_output function. 
I don't know why gcc is giving those warings. 

--- net/ipv4/route.c.old        2005-05-18 21:18:19.000000000 +0200
+++ net/ipv4/route.c    2005-05-28 08:23:18.000000000 +0200
@@ -1767,7 +1767,7 @@ static inline int ip_mkroute_input_def(s
                                       struct in_device *in_dev,
                                       u32 daddr, u32 saddr, u32 tos)
 {
-       struct rtable* rth;
+       struct rtable* rth = NULL;
        int err;
        unsigned hash;

@@ -2267,7 +2267,7 @@ static inline int ip_mkroute_output(stru
        unsigned char hop;
        unsigned hash;
        int err = -EINVAL;
-       struct rtable *rth;
+       struct rtable *rth = NULL;

        if (res->fi && res->fi->fib_nhs > 1) {
                unsigned char hopcount = res->fi->fib_nhs;
_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
http://lists.osdl.org/mailman/listinfo/kernel-janitors

Re: [KJ] GCC4 cleanup question

[Ricardo Nabinger Sanchez]

[-- Attachment #1: Type: text/plain, Size: 685 bytes --]

Quoting  Bongani Hlope <bonganilinux@mweb.co.za>
Sent on  Sat, 28 May 2005 08:25:54 +0200

> I don't have GCC4, the first rth is initialised inside the
> __mkroute_input function and the second one is initialised in the
> __mkroute_output function. I don't know why gcc is giving those warings. 

My guess is because it doesn't know/see that.  Wouldn't that be related to
the new interprocedural optimization?

More info, perhaps starting on:

	http://gcc.gnu.org/wiki/Interprocedural%20optimizations

-- 
Ricardo Nabinger Sanchez
GNU/Linux #140696 [http://counter.li.org]
Slackware Linux + FreeBSD

    How long a minute is depends on which 
    side of the bathroom door you're on.


[-- Attachment #2: Type: text/plain, Size: 167 bytes --]

_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
http://lists.osdl.org/mailman/listinfo/kernel-janitors

Re: [KJ] GCC4 cleanup question

[Arnd Bergmann]

On Sünnavend 28 Mai 2005 08:25, Bongani Hlope wrote:
> I don't have GCC4, the first rth is initialised inside the __mkroute_input 
> function and the second one is initialised in the __mkroute_output function. 
> I don't know why gcc is giving those warings. 

__mkroute_{in,out}put does not initialize *result if it returns early.

> --- net/ipv4/route.c.old        2005-05-18 21:18:19.000000000 +0200
> +++ net/ipv4/route.c    2005-05-28 08:23:18.000000000 +0200
> @@ -1767,7 +1767,7 @@ static inline int ip_mkroute_input_def(s
>                                        struct in_device *in_dev,
>                                        u32 daddr, u32 saddr, u32 tos)
>  {
> -       struct rtable* rth;
> +       struct rtable* rth = NULL;
>         int err;
>         unsigned hash;
> 

We've already had that discussion. Don't initialize it to a bogus value
in the declaration, but make sure it always gets set in the function
that is responsible for initializing it (__mkroute_input in this case).

	Arnd <><

_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
http://lists.osdl.org/mailman/listinfo/kernel-janitors